Use this URL to cite or link to this record in EThOS: https://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.807528
Title: The application of intelligent systems to finance and business
Author: Viner, James
Awarding Body: University of London
Current Institution: University College London (University of London)
Date of Award: 1999
Availability of Full Text:
Access from EThOS:
Full text unavailable from EThOS. Please try the link below.
Access from Institution:
Abstract:
Intelligent systems have proved very effective in many business applications, however there is little understanding of the relationships between the techniques and the application domains. Through applying genetic algorithms and neural networks, two powerful general purpose techniques to the difficult problems of economic forecasting and financial trading, this thesis investigates the connections between the nature of the application and the chosen intelligent technique. Four experiments have been carried out to investigate this problem: Residual Value Forecasting with Lex Vehicle Leasing: One method of setting vehicle hire charges requires accurate forecasts of the second hand value of vehicles in 3 or 4 years time. A synthetic depreciation series was constructed for proof-of-concept purposes. Neural network forecasting models were compared against linear regression benchmarks and it was found that they have comparable performance. Developments to this forecasting scheme are proposed. Intelligent Trade Filtering with Sabre Fund Management: This project is an attempt to capture, reproduce and extend expert trading knowledge from a history of pattern based trading. Genetic algorithms were used to find rules that capture the symbolic relationships between the observed market state (pattern type, quality etc.) at trade entry and probable trade outcome. The GA found several rules that are deterministic to 95% confidence. Continually Adaptive Trading Systems Design: Financial markets are in a constant state of change. Genetic algorithm-style operators were used on a population of trading system descriptions to generate new trading strategies on-the-fly which are evaluated on a rolling basis by their recent trading performance. In conjunction with the over-night loan market, the system could make super-LIBOR returns, although the impact of this result on theories of market efficiency is unclear. This system is unable to trade the FTSE index effectively, an observation consistent with the theory that the information set of the FTSE is too large for it to be out-performed. Genetic Algorithm Trading System Induction with the European Bank for Reconstruction and Development, a bank that speculatively trades government bond futures markets. Genetic algorithm rule induction was used to automate trading system innovation and profitability tests were carried out in all relevant markets. The results are positive but mixed. The system makes higher returns on longer maturity markets, and the presence of this effect over the US Treasury bond markets is demonstrated to a confidence of 86%. The system was also tested on copper and gold futures. The system was then modified to give a new technique for assessing the change in the character of financial markets. The principal scientific contributions to come from this thesis are: i) genetic algorithm rule induction is a powerful and effective technique for finding empirical models. It is particularly suitable for business problems as the experimenter has control over the representation and the resulting models are transparent; ii) novel results from the financial experiments present evidence both for, but primarily against, the Efficient Market Hypothesis; iii) that the decision to use a specific technology should be taken after the content of the available data has been investigated; iv) the proposition of a new technique for tracking changes in nature of financial markets using the trading rule induction engine; v) the design and operational analysis is given for the first known continuously adaptive trading engine; vi) the first known comprehensive operational analysis of a GA rule induction based trading system; vii) the first known public domain data intensive analysis of the vehicle resale market.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.807528  DOI: Not available
Share: